1. Data and methods

The data sources and methods of analysis in this study are presented briefly below and detailed in Annexes and in the methods article (da Corta et al., 2021, forthcoming). Throughout, we distinguish the experiences of youth on different poverty trajectories (see Box1). In our analysis, youth are generally defined as individuals aged between 18 and 35years old. This means that our sample mainly comprises individuals for whom (often new) work and marriage were the key life cycle stages.

Box 1. Poverty trajectories referenced in this study

• Chronic povertyis long-term poverty that persists over many years, or even a lifetime, and which is often transmitted inter-generationally, and this is how it is viewed in the qualitative data. In the quantitative data, it refers to households that are poor in both survey years. • Impoverishmentrefers to a poor person or household becoming poorer, or somebody who is nonpoor slipping into poverty. In the quantitative data, it refers to households who were non-poor in the 2011 survey and poor in 2014. In the qualitative data, we examine both impoverishment and transitory poverty escapes, and we also look at the distance moved into poverty (see Annex C for wellbeing groups). • Poverty escapes refers to households that escape poverty over time. In the panel data, this refers tohouseholds who were poor in 2011 and non-poor in 2014. In the qualitative data, we are also able to explore sustained poverty escapes (i.e. escapes sustained for five years). In the qualitative data, we also look at the distance moved away from poverty, with three wellbeing levels within the ‘poor’ category and three levels within the ‘non-poor’ category (see Annex C for more details). • Never poor refers to households who never fell into poverty during their lifetime. In the panel analysis, consists of households that were non-poor in both survey years.

Source: Adapted from Shepherd et al. (2014).

Quantitative methods

This study analyses the Living Standards Measurement Survey (LSMS) (L’Enquête nationale sur les conditions de vie des ménages et de l’agriculture, or ECVM/A), a nationally representative longitudinal survey of 3,436 households in rural and urban areas of Niger that took place in 2011 and 2014. Descriptive analysis of survey data is accompanied by multinomial logistic regressions to investigate drivers of different poverty trajectories across all households, and also restricted to the subset of youth-headed households. Auxiliary regressions are also conducted at the individual level to investigate correlates of salaried employment of young adults. In the survey, salaried employment is defined as the professional categories of ‘superior executive’, ‘middle executive’ or ‘master agent’, ‘qualified worker or employee’, ‘non-qualified worker or employee’, or ‘labourer’. We disaggregate these into executive agents and qualified workers in one group, and non-qualified workers or labourers as a separate group. Executives and qualified salaried employment are chosen

for analytic ease as proxies for resilient youth inclusion in labour markets, given wider literature pointing to the importance of salaried, stable employment in pathways out of poverty (Diwakar and Shepherd, 2018). However, while certain forms of salaried work re better paid than day work, private salaried work by month is not necessarily stable, as discussed in the findings. The models rely on baseline values of household (characteristics of head, assets, livelihood) and area/regional regressors, and shocks from the latest survey wave. All variables, summary statistics and regression results are presented in Annex D.

Qualitative methods

The Enquête Collective Rapide d'Identification des conflits et des groupes Stratégiques (ECRIS) framework used by Laboratoire d'Etudes et de Recherche sur les Dynamiques Sociales et le Développement Local (LASDEL) offers an opportunity for researchers and other actors involved as research stakeholders to jointly develop work on a common problematic and appropriate it. The ECRIS usually comprises of: (1) a preparatory phase, based on a benchmark survey, to identify target groups (also called strategic groups) and local issues related to the research themes; (2) a theoretical phase to clarify the ECRIS framework with its participants so as to collectively determine the target groups and identify provisional qualitative indicators for the surveys; and (3) a practical phase, alternating between surveying the sites targeted during the preparatory phase and group sessions for analysing and summarising the data collected. The ECRIS took place in Niamey and surveys were carried out over two days. We identified five strategic groups: (1) actors from NGOs, associations and unions; (2) teachers/trainers; (3) members of communities, such as local administrative and customary authorities or young people; (4) young employees and their employers; and (5) institutional (state and non-state) public and private actors, and technical and financial partners. Overall, 27 interviews were carried out as well as systematic observations in the survey sites. The Chronic Poverty Advisory Network (CPAN) method was sequenced in each of the four sites as follows.

1. Group discussions with knowledgeable long-term residents on key events and systemic changes. 2. Focus group discussions (FGDs) with men and women (separately), with equal numbers of young adults and older adults. These comprised of 16 group discussions of approximately 250 participants in total, and were designed to gather local perspectives on: (a) wellbeing (six rich-to-poor groups, with respondents from a range of socioeconomic statuses and occupations); (b) gender and generational social relations; and (c) Domestic abuse and disputes adjudication.

3. Youth life histories, sampled from the FGDs, for youth on different trajectories. Overall, 49 life histories, balanced between young men and women, were conducted and researchers were encouraged to identify pathways and processes underlying movements in and out of employment. 4. Key informant interviews (KIs) designed to address questions following life histories and group discussions: (a) relational KIs with people influential in the life of the youth (spouses, parents, trainers, mentors, employers, creditors); and (b) district-level and national KIs.

Mixedmethods data integration and analysis

A key goal of this paper is to integrate some of the LASDEL and CPAN qualitative methodologies in order to examine the access of young people to employment as a catalyst for their social inclusion. A point of integration was the ‘axis’ of the ECRIS approach –the focus on conflicts between groups as a ‘window’ or gateway to understanding youth inclusion (Bierschenk and Olivier de Sardan, 2014). The combined CPAN/LASDEL approach centralises conflict and uncovers ‘invisible’ social and bureaucratic norms as key causal drivers of the quality of youth inclusion in labour markets, as well as key moments in poverty pathways when young people negotiate with those more powerful than them and which enable or constrain further movement. The qualitative fieldwork in peri-urban and rural Zinder and Tahoua regions was conducted in 2020. The two rounds of field visits (January and March) involved 18 FGDs, 49 life history interviews (LHIs) and 4 KIs, which were used to analyse the evolution of youth inclusion in these regions of Niger.

Both the quantitative data and qualitative data are triangulated and build on the wider literature on poverty dynamics in Niger to contextualise the study findings. The qualitative data pays particular attention to panel survey years, while exploring in more detail the processes and trajectories of poverty over the respondent’s lifetime. As the qualitative data extends beyond the period covered by the survey, an escape from poverty in the quantitative datacould turn into either a sustained or transitory escape from poverty in the qualitative data. The analysis was also sequenced, with preliminary analysis of panel data offering insights to unpack in more detail through the qualitative investigations, afterwhich a more thorough quantitative analysis of the panel dataset was conducted that sought to build on the qualitative data findings.